Contributions of radar data for the estimation of biophysical parameters of agricultural surfaces.

Summary The work of this thesis is part of the South-West project, whose main objective is to contribute to the understanding and modeling of the functioning of continental surfaces at the landscape scale. This work aims to improve the monitoring and analysis capacities of highly anthropized surfaces: agrosystems. Both actors and spectators of climate change, these surfaces are also dedicated to food production. In this context, synthetic aperture radar (SAR) embedded in satellites has the double advantage of being sensitive to different parameters of continental surfaces (in relation to the soil or vegetation), and the ability to observe in cloudy conditions (unlike sensors operating in visible light). Since the 1990s, various studies based on images acquired with SAR technology have shown the interest of microwave data for monitoring continental surfaces. In recent years, the emergence of satellite missions in the X and L frequency bands has enriched the study possibilities previously limited to the C band. These sensor-satellite pairs now provide high spatial resolution products (up to one meter), with weekly revisit possibilities, necessary criteria for monitoring heterogeneous areas associated with strong temporal dynamicsThe work carried out in this thesis aims to establish the complementarity between radar (TerraSAR-X, Radarsat-2 and Alos, in the X, C and L spectral bands) and optical (Formosat-2, Spot-4/5) data acquired by satellites for monitoring agrosystems. The first is the implementation of an experimental campaign based on the acquisition of a set of data (satellite and field), necessary for the development of new approaches for landscape analysis. The monitored area, characterized by a strong anthropization, is located 50 km southwest of Toulouse. The satellite images include three radar time series (X, C and L bands), plus optical acquisitions (Formosat-2, Spot-4/5). With a total of a hundred images acquired in the microwave, the area common to the various scenes covers an area of 10×10 km ². At the same time, the field measurement protocols have made it possible to consider independently the two key elements of the surface: the soil and the crop. In addition to the meteorological stations installed within the framework of the site, qualitative and quantitative measurements were carried out in a synchronous way with the satellite acquisitions, on a total of 387 plots. The temporal signatures of each crop were then established at each satellite acquisition wavelength (optical and radar) using an original approach of angular normalization of radar signals (combination of radar and optical information). The results obtained during the phenological cycle of winter crops (wheat and rapeseed) and summer crops (maize, soybean and sunflower) clearly show the complementarity of the multi-sensor approaches, and the specificity of the radar signals (in relation to the polarization states and frequencies considered). Two biophysical parameters related to the vegetation are finally estimated (LAI and height), the microwave data showing both an important sensitivity and good performances.1 The electromagnetic modelling on bare soil has first allowed to evaluate different formalisms, namely: the models of Dubois and Oh (1992 and 2004) having as common characteristics a simplified description of the processes. They are confronted with a model based on physical foundations, the IEM model (Integral Equation Model). The application of the models in the different spectral bands (X, C and L), shows very heterogeneous results, the best performances being obtained in X band, with the model of Oh 1992. Subsequently, the improvement of the models takes advantage of the analysis of the residuals (with respect to the input variables), in order to reduce the observed dispersion. The models tested are optimized and validated using a residuals approach. The results highlight the interest of multi-sensor data for the monitoring of surfaces dedicated to agriculture. In the near future, space missions such as Tandem-X, Sentinel-1/-2, Radarsat Constellation or Alos-2 should provide access to these data, and thus clarify the results obtained in this thesis.